Tesseract OCR tsv to json converter

tesseract-tsv-to-json.js

const tesseractTsvToJson = (tsvFilename, jsonFilename) => {
    console.log(`> tsvToJson ${tsvFilename} => ${jsonFilename}`);
    const tsvLines = fs.readFileSync(tsvFilename, 'utf-8').trim().split('\n').reverse();

    /** @type {any} */
    const json = {pages: []};

    // discard header
    tsvLines.pop();

    while (tsvLines.length) {
        let [
            level,
            pageNumberRaw,
            blockNumberRaw,
            paragraphNumberRaw,
            lineNumberRaw,
            wordNumberRaw,
            left,
            top,
            width,
            height,
            confidence,
            text = '',
        ] = tsvLines.pop()?.trim().split('\t') ?? [];

        const pageNumber = Number(pageNumberRaw) - 1;
        const blockNumber = Number(blockNumberRaw) - 1;
        const paragraphNumber = Number(paragraphNumberRaw) - 1;
        const lineNumber = Number(lineNumberRaw) - 1;
        const wordNumber = Number(wordNumberRaw) - 1;

        const box = {top, left, width, height};

        if (!json.pages[pageNumber]) {
            json.pages[pageNumber] = {blocks: [], ...box};
            continue;
        }
        const page = json.pages[pageNumber];

        if (!page.blocks[blockNumber]) {
            page.blocks[blockNumber] = {paragraphs: [], ...box};
            continue;
        }
        const block = page.blocks[blockNumber];

        if (!block.paragraphs[paragraphNumber]) {
            block.paragraphs[paragraphNumber] = {lines: [], ...box};
            continue;
        }
        const paragraph = block.paragraphs[paragraphNumber];

        if (!paragraph.lines[lineNumber]) {
            paragraph.lines[lineNumber] = {words: [], ...box};
            continue;
        }
        const line = paragraph.lines[lineNumber];

        if (!line.words[wordNumber]) {
            line.words[wordNumber] = {text, ...box};
        }
    }
    fs.writeFileSync(jsonFilename, JSON.stringify(json, null, 2));
};